Skip to content
neuroflow0.2.31 · docs

← neuroflow

phase-grant-proposal

The grant-proposal phase translates a defined research question into a competitive, funder-ready application. It begins with automatic discovery of ideation outputs and ends with a fully drafted, word-counted, reviewer-aligned document saved to .neuroflow/grant-proposal/.

Phase entry checklist

Before drafting anything, verify:

  • Researcher interview complete — answers saved to interview-[funder]-[date].md
  • Objectives written to .neuroflow/objectives.md (one numbered sentence per objective)
  • Ideation output discovered OR research question confirmed with user
  • Funder and scheme identified
  • Page/word limits per section confirmed
  • Review criteria retrieved (from funder website or call document)
  • Deadline confirmed and logged

Never skip the interview. Never draft without confirmed objectives.

Interview protocol

Ask questions one or two at a time — conversational, not a form. Build context progressively.

Required questions: 1. Core research question (offer ideation summary if .neuroflow/ideation/ exists) 2. Funder and scheme (or URL to the call) 3. Budget ceiling and duration 4. Deadline 5. Specific objectives or aims — ask for 3–4 verb-led statements; these are the cornerstones 6. Preliminary data — what exists and how strong is the feasibility case? 7. Novelty — what does this do that current methods don't? 8. Team — PI and co-investigators with roles 9. Constraints — ethics status, required partnerships, pre-committed budget items 10. Previous grants for inspiration — if provided, build an inspiration map

After the interview: - Save to .neuroflow/grant-proposal/interview-[funder]-[date].md - Write objectives to .neuroflow/objectives.md (create or update)

Objectives tracking

Objectives in .neuroflow/objectives.md are cornerstones for the entire grant. Rules:

  1. Re-read objectives.md before drafting every section
  2. For the Approach/Methodology section: verify all N objectives appear explicitly — flag any missing
  3. For the quality checklist: verify all N objectives appear in both Aims AND Methodology
  4. If the user adds or changes an objective mid-session: update objectives.md immediately and note the change in reasoning/grant-proposal.jsonl

Panel research protocol

When the user agrees to panel research:

  1. Ask for the panel listing URL
  2. Fetch the page (WebFetch)
  3. For each panel member found: search their background (WebSearch — 2–3 key papers, primary topics, methodological focus)
  4. Build a panel profile table: | Panel name | Member | Research background | Relevance to project |
  5. Suggest top 2–3 panels with rationale (one sentence each)
  6. Save to .neuroflow/grant-proposal/panels/panel-analysis-[funder]-[date].md
  7. Use the preferred panel's terminology and methodological frame throughout the proposal — reviewers respond to language that mirrors their own thinking

Inspiration map protocol

When the user provides previous grant documents:

  1. Read each document in full
  2. Build a cross-reference table:
| New grant section   | Grant A              | Grant B              |
|---------------------|----------------------|----------------------|
| Specific Aims       | Sec 1 (pp. 1–2)      | Introduction (pp. 1–3)|
| Background          | Lit Review (pp. 3–6) | Background (pp. 2–5) |
| Innovation          | Novelty (p. 7)       | Not present          |
| Approach            | Methods (pp. 8–14)   | Research Plan (pp. 6–12)|
| Budget              | Budget (pp. 15–16)   | Budget narrative (p.13)|
  1. Save to .neuroflow/grant-proposal/inspiration-map-[date].md
  2. Before drafting each section: reference the corresponding inspiration entry — adopt the structural approach and framing style while producing original content
  3. When inspiration sections contain strong language or argument patterns: note them explicitly in the draft as structural guides

Ideation discovery protocol

If ideation files exist (.neuroflow/ideation/*.md):

  1. List all files
  2. Read each and extract: research question, modality, population, preliminary data, key references
  3. Present a 3–5 bullet summary during the interview
  4. Ask: "Is this still the right question to build the grant around?"

If no ideation files exist: the interview questions cover everything needed.

Funder knowledge base

Budget figures, page limits, and scheme names are approximate typical values as of 2024–2025. Verify against the current call document or funder website before advising a user — requirements change each cycle.

NIH (USA)

  • Mechanisms: R01 (5 yr, $500K DC/yr), R21 (2 yr, $275K total), R03 (2 yr, $50K/yr), K awards (career development), U01 (cooperative agreement)
  • Key sections: Specific Aims (1 page — the most critical), Research Strategy (12p for R01: Significance, Innovation, Approach), Human Subjects, Authentication of Resources, Bibliography
  • Review criteria (scored 1–9, lower = better): Significance (gap and impact), Investigators (team track record), Innovation (conceptual or methodological novelty), Approach (rigor, feasibility, power analysis), Environment (facilities, collaborations)
  • Common weaknesses cited by study sections: insufficient power analysis, vague alternatives for failed aims, overambitious scope, lack of preliminary data, budget not justified
  • Register on: NIH eRA Commons; submit via Grants.gov or ASSIST

ERC (EU)

  • Mechanisms: Starting Grant (≤7 yr post-PhD, €1.5M), Consolidator (7–12 yr post-PhD, €2M), Advanced Grant (established PIs, €2.5M), Synergy Grant (2–4 PIs, €10M)
  • Key sections: B1 Extended Synopsis (5p), B2 Full Proposal (Part B1 + B2 up to 15p for StG; 50p for AdG)
  • Review criteria: Scientific excellence (primary), Impact (to science and society), Quality and efficiency of implementation (team, resources, management)
  • Common weaknesses: proposal not ambitious enough for ERC standards, host institution not sufficiently described, no clear "frontier research" framing
  • Note: ERC funds PI-driven curiosity research, not translational or applied work — frame accordingly

Wellcome Trust (UK)

  • Mechanisms: Discovery Award (£3–5M, 5 yr), Investigator Award (£1–3M, 5 yr), Sir Henry Wellcome (4 yr postdoc award), Collaborative Award
  • Key sections: Flexible — Wellcome provides a template; typically Summary, Scientific rationale, Approach, Impact, Team, Budget
  • Review criteria: Scientific opportunity (is this the right question?), Team (can they deliver?), Delivery (is the plan achievable?)
  • Note: Wellcome values interdisciplinary approaches and expects explicit attention to open science (data sharing, preregistration)

National research councils and science foundations

  • Mechanisms: most countries run a research council or science foundation with standard project grants (often 3 years), junior and excellence schemes, programme grants, and bilateral calls with partner countries
  • Key sections: usually a project summary (often with a strict word limit), state of the art, objectives and hypotheses, methodology, feasibility, requested funds and justification, timeline, and short CVs — the call text and the template decide
  • Review criteria: usually originality and scientific quality, feasibility, team qualification, and budget adequacy — take them from the call and map each to a section (Review criteria alignment checklist below)
  • Note: check the call text for language and format requirements. National funders often want some parts (the summary, sometimes the whole application) in the national language and others in English for international reviewers; templates, fonts and page rules are strict, and an application that breaks them can be rejected unread

Horizon Europe (EU)

  • Mechanisms: EIC Pathfinder (€3M, 4 yr), EIC Transition, EIC Accelerator (SMEs), MSCA Fellowships, ERC (separate)
  • Key sections: Excellence (concept, methodology), Impact (strategy, pathways to impact), Implementation (timeline, resources, team)
  • Review criteria: Excellence (novelty, scientific quality), Impact (significance, dissemination), Implementation (coherence, resources)

Neuroscience-specific grant writing tactics

Power analysis

  • Always include a formal power analysis, even if approximate
  • For EEG/ERP: reference pilot data or published effect sizes (Cohen's d for amplitude differences; cite the source)
  • For fMRI: cite comparable studies for BOLD effect sizes; state voxel-wise threshold + cluster correction
  • For behavioural measures: G*Power output with exact parameters (α, 1-β, effect size, allocation ratio)
  • Reviewers from biostatistics panels will look for this — a missing power analysis is a common scoring weakness

Preprocessing and analysis plan

  • Name the software pipeline explicitly (MNE-Python, EEGLAB, SPM12, FSL, Brainstorm, ERPLAB)
  • State: sampling rate, bandpass filter cutoffs, epoch window, artifact rejection method (ICA, threshold)
  • For fMRI: TR, slice timing, HRF model, motion scrubbing threshold
  • Reference BIDS compliance if relevant

Participant recruitment

  • State inclusion/exclusion criteria precisely
  • Address any vulnerable populations explicitly (IRB / ethics committee considerations)
  • If EEG: mention net type, cap system, electrode count
  • Provide realistic recruitment numbers with dropout correction (plan for 20% attrition minimum)

Preliminary data

  • Show it early — link pilot data to feasibility
  • If no preliminary data exists: substitute published proof-of-concept studies with your team's commentary
  • Use figures — a single figure showing pilot ERP or BOLD effect is worth 200 words

Budget granularity

  • Break personnel costs to FTE fractions (PI 20%, postdoc 100%, RA 50%, etc.)
  • List equipment by model number and price if >$5K items
  • Consumables: EEG gel/electrode replacement, MRI contrast agents, participant payment
  • Travel: conference attendance, site visits (justify each)
  • Indirect costs: confirm the correct rate for your institution

Review criteria alignment checklist

For each major funder, map every section of the proposal to the review criterion it addresses:

Section NIH criterion ERC criterion Wellcome criterion
Specific Aims Significance + Innovation Scientific excellence Scientific opportunity
Background Significance Scientific excellence Scientific opportunity
Innovation Innovation Scientific excellence —
Approach Approach + Environment Quality of implementation Delivery
Team Investigators Quality of implementation Team
Budget — Quality of implementation Delivery
Preliminary data Approach Scientific excellence Delivery

Use this mapping to ensure reviewer criteria are explicitly addressed in the text — never assume reviewers will infer alignment.

Common fatal weaknesses (avoid these)

  1. Overambitious scope — three aims that cannot be done in the funded period
  2. No power analysis — instant credibility loss with quantitative reviewers
  3. Vague alternatives — "if Aim 1 fails, we will modify our approach" is not acceptable
  4. Hypothesis-free Approach — methods with no testable prediction attached
  5. Budget not justified — line items without rationale
  6. Ignoring funder priorities — NIH wants disease relevance; ERC wants frontier science; these are different framings
  7. Insufficient preliminary data — especially for NIH R01; reviewers score feasibility based on prior work
  8. Poor Specific Aims page — the single most influential page in a US grant; if reviewers are not sold after 1 page, scores suffer

Workflow integration

Reads from

  • .neuroflow/ideation/ — research question, literature, study design drafts
  • .neuroflow/project_config.md — team, institution, phase, modality
  • .neuroflow/objectives.md — project objectives cornerstones (re-read before every section)
  • .neuroflow/grant-proposal/flow.md — prior grant context if resuming
  • Previous grant documents (provided by user) — for inspiration map

Writes to

  • .neuroflow/objectives.md — written/updated after interview confirms objectives
  • .neuroflow/grant-proposal/interview-[funder]-[date].md — interview transcript
  • .neuroflow/grant-proposal/inspiration-map-[date].md — cross-reference table (if previous grants provided)
  • .neuroflow/grant-proposal/panels/panel-analysis-[funder]-[date].md — panel profiles (if panel research done)
  • .neuroflow/grant-proposal/draft-[funder]-[date]-[section].md — per-section drafts
  • .neuroflow/grant-proposal/grant-[funder]-[date].md — full consolidated draft
  • .neuroflow/grant-proposal/ai-use-statement-[funder]-[date].md — AI-use declaration, drafted from the session logs and confirmed by the person
  • .neuroflow/grant-proposal/flow.md — funder, scheme, deadline, section status
  • .neuroflow/reasoning/grant-proposal.jsonl — funder and scheme choice, objective changes
  • .neuroflow/sessions/YYYY-MM-DD.md — session log (using ## milestone headers)
  • .neuroflow/timeline.md — the deadline; project_config.md — funder and scheme as notes below the frontmatter (with user confirmation)

Connects to

  • /ideation — if no research question exists, redirect the user there first
  • /experiment — paradigm details flow into the Approach section
  • /preregistration — registered analysis plan can be referenced in the Approach
  • /write-report — funder progress reports use the same structure

Output format standard

Each drafted section is presented as a standalone block:

## [Section name]  ([page limit] page / [word limit] words)

[Drafted content]

---
Word count: NNN / NNN limit    ✅ within limit  /  ⚠️ over by NNN words

After each section, offer: - "revise" — rework the current section - "next" — proceed to the next section - "expand [topic]" — add depth to a specific part of the current section - "save" — write current section to the draft file - "checklist" — run the quality checklist against the current section

Relevant skills

  • neuroflow:neuroflow-core — lifecycle and .neuroflow/ write rules (read first)
  • neuroflow:phase-ideation — if user needs to define the research question first
  • neuroflow:phase-experiment — paradigm design details for the Approach section
  • neuroflow:phase-preregistration — link registered analysis plan
  • neuroflow:humanizer — optional style edit (filler, rhythm, register), only when the person asks for it; never automatic and never used to disguise AI involvement — AI use is declared (command Step 6b)
  • sequentialthinking MCP — invoke before drafting Innovation and Approach sections to build a rigorous logical argument chain

Slash command

/neuroflow:grant-proposal — runs this workflow as a slash command.